Rethinking the Reranker: Boundary-Aware Evidence Selection for Robust Retrieval-Augmented Generation
Published in International Conference on Machine Learning (ICML 2026), 2026
Recommended citation: Jiashuo Sun, Pengcheng Jiang, Saizhuo Wang, Jiajun Fan, Heng Wang, Siru Ouyang, Ming Zhong, Yizhu Jiao, Chengsong Huang, Xueqiang Xu, Pengrui Han, Peiran Li, Jiaxin Huang, Ge Liu, Heng Ji, Jiawei Han. "Rethinking the Reranker: Boundary-Aware Evidence Selection for Robust Retrieval-Augmented Generation." ICML 2026. https://openreview.net/forum?id=Tt8lCe1NrW
BAR-RAG reframes the reranker as a boundary-aware evidence selector that targets the generator’s “Goldilocks Zone” — evidence that is neither trivially answer-revealing nor unanswerable, but challenging yet sufficient. The selector is trained with RL from generator feedback, then the generator is fine-tuned under the induced evidence distribution. Average gain of 10.3% over strong RAG and reranking baselines under noisy retrieval. Code.